!unzip -q "archive.zip"

import os

import numpy as np
import pandas as pd

import cv2
import matplotlib.pyplot as plt
import seaborn as sns

%matplotlib inline

base_dir = "/content/Skin cancer ISIC The International Skin Imaging Collaboration"
train_dir = os.path.join(base_dir, "Train")

records = []

for class_name in os.listdir(train_dir):
    class_folder = os.path.join(train_dir, class_name)

    for fname in os.listdir(class_folder):
        fpath = os.path.join(class_folder, fname)
        records.append({
            "filepath": fpath,
            "label": class_name
        })

train_df = pd.DataFrame(records)

train_df.head()

print("Total number of images in the training set:", len(train_df))
print("\nNumber of images per class:")
print(train_df["label"].value_counts())

Total number of images in the training set: 2239

Number of images per class:
label
pigmented benign keratosis    462
melanoma                      438
basal cell carcinoma          376
nevus                         357
squamous cell carcinoma       181
vascular lesion               139
actinic keratosis             114
dermatofibroma                 95
seborrheic keratosis           77
Name: count, dtype: int64

heights = []
widths = []
channels = []
filesizes = []

for path in train_df["filepath"]:
    img = cv2.imread(path)
    if img is None:
        heights.append(np.nan)
        widths.append(np.nan)
        channels.append(np.nan)
        filesizes.append(np.nan)
        continue

    h, w = img.shape[:2]
    c = 1 if img.ndim == 2 else img.shape[2]

    heights.append(h)
    widths.append(w)
    channels.append(c)

    filesizes.append(os.path.getsize(path))

train_df["height"] = heights
train_df["width"] = widths
train_df["channels"] = channels
train_df["filesize_bytes"] = filesizes

train_df.head()

print(train_df["channels"].value_counts())

channels
3    2239
Name: count, dtype: int64

plt.figure(figsize=(6,4))
plt.hist(train_df["filesize_bytes"] / 1024, bins=30)
plt.xlabel("File size (KB)")
plt.ylabel("Number of images")
plt.title("Distribution of image file sizes")
plt.tight_layout()
plt.show()

sample_df = train_df.sample(9, random_state=42).reset_index(drop=True)

plt.figure(figsize=(10, 18))

for i, row in sample_df.iterrows():
    img_bgr = cv2.imread(row["filepath"])
    img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)
    img_gray = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY)

    # RGB image
    plt.subplot(9, 2, 2*i + 1)
    plt.imshow(img_rgb)
    plt.axis("off")
    plt.title(f"RGB - {row['label']}", fontsize=8)

    # Grayscale image
    plt.subplot(9, 2, 2*i + 2)
    plt.imshow(img_gray, cmap="gray")
    plt.axis("off")
    plt.title("Grayscale", fontsize=8)

plt.tight_layout()
plt.show()

stats = []

for i, row in sample_df.iterrows():
    img_bgr = cv2.imread(row["filepath"])
    img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)
    img_gray = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY)

    stats.append({
        "filepath": row["filepath"],
        "label": row["label"],

        # RGB statistics (combined across channels)
        "rgb_min":  img_rgb.min(),
        "rgb_max":  img_rgb.max(),
        "rgb_mean": img_rgb.mean(),
        "rgb_std":  img_rgb.std(),

        # Grayscale statistics
        "gray_min":  img_gray.min(),
        "gray_max":  img_gray.max(),
        "gray_mean": img_gray.mean(),
        "gray_std":  img_gray.std()
    })

stats_df = pd.DataFrame(stats)
stats_df

plt.figure(figsize=(12, 35))

for i, row in sample_df.iterrows():
    img_bgr = cv2.imread(row["filepath"])
    img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)
    img_gray = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY)

    # --- RGB HISTOGRAM ---
    plt.subplot(9, 2, 2*i + 1)

    # Red channel
    plt.hist(img_rgb[:,:,0].ravel(), bins=256, color='red', alpha=0.5, label='Red')
    # Green channel
    plt.hist(img_rgb[:,:,1].ravel(), bins=256, color='green', alpha=0.5, label='Green')
    # Blue channel
    plt.hist(img_rgb[:,:,2].ravel(), bins=256, color='blue', alpha=0.5, label='Blue')

    plt.title(f"RGB Histogram - {row['label']}", fontsize=9)
    plt.xlabel("Pixel Intensity")
    plt.ylabel("Frequency")
    plt.legend()

    # --- GRAYSCALE HISTOGRAM ---
    plt.subplot(9, 2, 2*i + 2)
    plt.hist(img_gray.ravel(), bins=256, color='gray')
    plt.title("Grayscale Histogram", fontsize=9)
    plt.xlabel("Pixel Intensity")
    plt.ylabel("Frequency")

plt.tight_layout()
plt.show()

row = sample_df.iloc[0]
img_bgr = cv2.imread(row["filepath"])
img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)
img_gray = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY)

def contrast_stretch(img):
    """Apply min-max contrast stretching to [0, 255]."""
    img = img.astype(np.float32)
    min_val = img.min()
    max_val = img.max()
    stretched = (img - min_val) * 255.0 / (max_val - min_val + 1e-8)
    return np.clip(stretched, 0, 255).astype(np.uint8)

# RGB: stretch each channel separately
rgb_stretched = np.zeros_like(img_rgb)
for c in range(3):
    rgb_stretched[:,:,c] = contrast_stretch(img_rgb[:,:,c])

# Grayscale stretching
gray_stretched = contrast_stretch(img_gray)

# --- Show results ---
plt.figure(figsize=(10,8))

plt.subplot(2,2,1)
plt.imshow(img_rgb)
plt.title(f"Original RGB - {row['label']}")
plt.axis("off")

plt.subplot(2,2,2)
plt.imshow(rgb_stretched)
plt.title("Contrast Stretched RGB")
plt.axis("off")

plt.subplot(2,2,3)
plt.imshow(img_gray, cmap="gray")
plt.title("Original Grayscale")
plt.axis("off")

plt.subplot(2,2,4)
plt.imshow(gray_stretched, cmap="gray")
plt.title("Contrast Stretched Grayscale")
plt.axis("off")

plt.tight_layout()
plt.show()

# --- RGB: Histogram Equalization on Y channel in YCrCb ---

img_ycrcb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2YCrCb)
y, cr, cb = cv2.split(img_ycrcb)

# equalize only Y
y_eq = cv2.equalizeHist(y)

img_ycrcb_eq = cv2.merge([y_eq, cr, cb])
img_bgr_eq = cv2.cvtColor(img_ycrcb_eq, cv2.COLOR_YCrCb2BGR)
img_rgb_eq = cv2.cvtColor(img_bgr_eq, cv2.COLOR_BGR2RGB)

# --- Grayscale equalization ---
gray_eq = cv2.equalizeHist(img_gray)

# Visualization
plt.figure(figsize=(10,8))

plt.subplot(2,2,1)
plt.imshow(img_rgb)
plt.title("Original RGB")
plt.axis("off")

plt.subplot(2,2,2)
plt.imshow(img_rgb_eq)
plt.title("Histogram Equalized RGB")
plt.axis("off")

plt.subplot(2,2,3)
plt.imshow(img_gray, cmap="gray")
plt.title("Original Grayscale")
plt.axis("off")

plt.subplot(2,2,4)
plt.imshow(gray_eq, cmap="gray")
plt.title("Histogram Equalized Grayscale")
plt.axis("off")

plt.tight_layout()
plt.show()

def gamma_correction(img, gamma):
    """Apply gamma correction with a given gamma value."""
    img_norm = img.astype(np.float32) / 255.0      # [0,1] aralığına çek
    corrected = np.power(img_norm, gamma)          # gamma uygula
    corrected = np.clip(corrected * 255.0, 0, 255) # geri [0,255]
    return corrected.astype(np.uint8)

gammas = [0.5, 1.0, 2.0]

# RGB gamma correction
rgb_gamma_images = [gamma_correction(img_rgb, g) for g in gammas]

# Grayscale gamma correction
gray_gamma_images = [gamma_correction(img_gray, g) for g in gammas]

# Show results
plt.figure(figsize=(12, 8))

# RGB row
for i, g in enumerate(gammas):
    plt.subplot(2, 3, i + 1)
    plt.imshow(rgb_gamma_images[i])
    plt.title(f"RGB - gamma = {g}")
    plt.axis("off")

# Grayscale row
for i, g in enumerate(gammas):
    plt.subplot(2, 3, i + 4)
    plt.imshow(gray_gamma_images[i], cmap="gray")
    plt.title(f"Grayscale - gamma = {g}")
    plt.axis("off")

plt.tight_layout()
plt.show()

# Median blur (k=5)
median_rgb = cv2.medianBlur(img_rgb, 5)
median_gray = cv2.medianBlur(img_gray, 5)

# Gaussian blur (k=5, sigmaX=0)
gaussian_rgb = cv2.GaussianBlur(img_rgb, (5, 5), 0)
gaussian_gray = cv2.GaussianBlur(img_gray, (5, 5), 0)

plt.figure(figsize=(12, 12))

# ---------------- RGB column ----------------
plt.subplot(3, 2, 1)
plt.imshow(img_rgb)
plt.title("Original RGB")
plt.axis("off")

plt.subplot(3, 2, 3)
plt.imshow(median_rgb)
plt.title("Median Blurred RGB")
plt.axis("off")

plt.subplot(3, 2, 5)
plt.imshow(gaussian_rgb)
plt.title("Gaussian Blurred RGB")
plt.axis("off")

# ---------------- Grayscale column ----------------
plt.subplot(3, 2, 2)
plt.imshow(img_gray, cmap="gray")
plt.title("Original Grayscale")
plt.axis("off")

plt.subplot(3, 2, 4)
plt.imshow(median_gray, cmap="gray")
plt.title("Median Blurred Grayscale")
plt.axis("off")

plt.subplot(3, 2, 6)
plt.imshow(gaussian_gray, cmap="gray")
plt.title("Gaussian Blurred Grayscale")
plt.axis("off")

plt.tight_layout()
plt.show()

angle = np.random.uniform(0, 10)
print("Random rotation angle:", angle)

h, w = img_rgb.shape[:2]
center = (w // 2, h // 2)

M = cv2.getRotationMatrix2D(center, angle, 1.0)

rot_rgb = cv2.warpAffine(img_rgb, M, (w, h))
rot_gray = cv2.warpAffine(img_gray, M, (w, h))

plt.figure(figsize=(10, 8))

plt.subplot(2, 2, 1)
plt.imshow(img_rgb)
plt.title("Original RGB")
plt.axis("off")

plt.subplot(2, 2, 2)
plt.imshow(rot_rgb)
plt.title(f"Rotated RGB ({angle:.2f} deg)")
plt.axis("off")

plt.subplot(2, 2, 3)
plt.imshow(img_gray, cmap="gray")
plt.title("Original Grayscale")
plt.axis("off")

plt.subplot(2, 2, 4)
plt.imshow(rot_gray, cmap="gray")
plt.title(f"Rotated Grayscale ({angle:.2f} deg)")
plt.axis("off")

plt.tight_layout()
plt.show()

Random rotation angle: 2.845367545878148

flip_rgb = cv2.flip(img_rgb, 1)
flip_gray = cv2.flip(img_gray, 1)

plt.figure(figsize=(10, 8))

plt.subplot(2, 2, 1)
plt.imshow(img_rgb)
plt.title("Original RGB")
plt.axis("off")

plt.subplot(2, 2, 2)
plt.imshow(flip_rgb)
plt.title("Horizontally Flipped RGB")
plt.axis("off")

plt.subplot(2, 2, 3)
plt.imshow(img_gray, cmap="gray")
plt.title("Original Grayscale")
plt.axis("off")

plt.subplot(2, 2, 4)
plt.imshow(flip_gray, cmap="gray")
plt.title("Horizontally Flipped Grayscale")
plt.axis("off")

plt.tight_layout()
plt.show()

gray = cv2.cvtColor(img_rgb, cv2.COLOR_RGB2GRAY)

# FFT
f = np.fft.fft2(gray)
fshift = np.fft.fftshift(f)
magnitude_spectrum = 20 * np.log(np.abs(fshift) + 1)

plt.figure(figsize=(12, 6))

plt.subplot(1, 2, 1)
plt.imshow(gray, cmap='gray')
plt.title("Grayscale Image")
plt.axis("off")

plt.subplot(1, 2, 2)
plt.imshow(magnitude_spectrum, cmap='gray')
plt.title("FFT Magnitude Spectrum")
plt.axis("off")

plt.tight_layout()
plt.show()

rows, cols = gray.shape
crow, ccol = rows // 2, cols // 2

# radius for low-pass filter (change radius for stronger/weaker blur)
radius = 40

# Create mask
mask = np.zeros((rows, cols), np.uint8)
cv2.circle(mask, (ccol, crow), radius, 1, -1)

# Apply mask in frequency domain
fshift_filtered = fshift * mask

magnitude_spectrum_filtered = 20 * np.log(np.abs(fshift_filtered) + 1)

plt.figure(figsize=(12, 6))

plt.subplot(1, 2, 1)
plt.imshow(mask, cmap='gray')
plt.title("Low-Pass Filter Mask")
plt.axis("off")

plt.subplot(1, 2, 2)
plt.imshow(magnitude_spectrum_filtered, cmap='gray')
plt.title("Filtered Magnitude Spectrum")
plt.axis("off")

plt.tight_layout()
plt.show()

# 6.3 – Inverse FFT
f_ishift = np.fft.ifftshift(fshift_filtered)
img_back = np.fft.ifft2(f_ishift)
img_back = np.abs(img_back)

# 6.4 – Karşılaştırma görseli: orijinal vs low-pass (FFT sonucu)
plt.figure(figsize=(12, 6))

plt.subplot(1, 2, 1)
plt.imshow(gray, cmap='gray')
plt.title("Original Grayscale")
plt.axis("off")

plt.subplot(1, 2, 2)
plt.imshow(img_back, cmap='gray')
plt.title("Reconstructed Image (Low-Pass FFT)")
plt.axis("off")

plt.tight_layout()
plt.show()

blur_kernel = (9, 9)
alpha = 1.5  # sharpening strength

# --- RGB sharpening ---
blur_rgb = cv2.GaussianBlur(img_rgb, blur_kernel, 0)
mask_rgb = cv2.subtract(img_rgb, blur_rgb)
sharp_rgb = cv2.addWeighted(img_rgb, 1 + alpha, blur_rgb, -alpha, 0)

# --- Grayscale sharpening ---
blur_gray = cv2.GaussianBlur(img_gray, blur_kernel, 0)
mask_gray = cv2.subtract(img_gray, blur_gray)
sharp_gray = cv2.addWeighted(img_gray, 1 + alpha, blur_gray, -alpha, 0)

plt.figure(figsize=(12, 10))

# RGB
plt.subplot(2, 3, 1)
plt.imshow(img_rgb)
plt.title("Original RGB")
plt.axis("off")

plt.subplot(2, 3, 2)
plt.imshow(blur_rgb)
plt.title("Blurred RGB")
plt.axis("off")

plt.subplot(2, 3, 3)
plt.imshow(sharp_rgb)
plt.title("Sharpened RGB (Unsharp Mask)")
plt.axis("off")

# Grayscale
plt.subplot(2, 3, 4)
plt.imshow(img_gray, cmap='gray')
plt.title("Original Grayscale")
plt.axis("off")

plt.subplot(2, 3, 5)
plt.imshow(blur_gray, cmap='gray')
plt.title("Blurred Grayscale")
plt.axis("off")

plt.subplot(2, 3, 6)
plt.imshow(sharp_gray, cmap='gray')
plt.title("Sharpened Grayscale (Unsharp Mask)")
plt.axis("off")

plt.tight_layout()
plt.show()

scale = 2.0

# Bicubic upscaling (INTER_CUBIC)
sharp_rgb_up = cv2.resize(
    sharp_rgb,
    None,
    fx=scale,
    fy=scale,
    interpolation=cv2.INTER_CUBIC
)

sharp_gray_up = cv2.resize(
    sharp_gray,
    None,
    fx=scale,
    fy=scale,
    interpolation=cv2.INTER_CUBIC
)

plt.figure(figsize=(10, 10))

plt.subplot(2, 1, 1)
plt.imshow(sharp_rgb_up)
plt.title("Sharpened RGB (Displayed in Upscaled Size)")
plt.axis("off")

plt.subplot(2, 1, 2)
plt.imshow(sharp_gray_up, cmap="gray")
plt.title("Sharpened Grayscale (Displayed in Upscaled Size)")
plt.axis("off")

plt.tight_layout()
plt.show()

	filepath	label	height	width	channels	filesize_bytes
0	/content/Skin cancer ISIC The International Sk...	basal cell carcinoma	450	600	3	270438
1	/content/Skin cancer ISIC The International Sk...	basal cell carcinoma	450	600	3	219575
2	/content/Skin cancer ISIC The International Sk...	basal cell carcinoma	450	600	3	331292
3	/content/Skin cancer ISIC The International Sk...	basal cell carcinoma	450	600	3	148557
4	/content/Skin cancer ISIC The International Sk...	basal cell carcinoma	450	600	3	146123

	filepath	label	rgb_min	rgb_max	rgb_mean	rgb_std	gray_min	gray_max	gray_mean	gray_std
0	/content/Skin cancer ISIC The International Sk...	dermatofibroma	62	255	181.391322	31.522716	98	230	176.884252	21.455418
1	/content/Skin cancer ISIC The International Sk...	basal cell carcinoma	1	244	180.642422	30.416353	13	223	180.061519	20.946992
2	/content/Skin cancer ISIC The International Sk...	pigmented benign keratosis	0	217	147.761711	24.677234	4	198	142.397696	21.637895
3	/content/Skin cancer ISIC The International Sk...	nevus	9	255	179.213149	43.150476	44	252	180.824000	32.811327
4	/content/Skin cancer ISIC The International Sk...	pigmented benign keratosis	0	229	117.074749	22.531714	1	186	112.179004	18.360078
5	/content/Skin cancer ISIC The International Sk...	dermatofibroma	60	244	153.870544	44.219801	95	198	148.719215	19.560209
6	/content/Skin cancer ISIC The International Sk...	nevus	12	188	152.545577	26.667685	30	178	152.702267	23.470868
7	/content/Skin cancer ISIC The International Sk...	nevus	0	255	187.230696	49.002421	50	255	187.244869	19.775074
8	/content/Skin cancer ISIC The International Sk...	nevus	11	255	164.373313	33.599677	19	230	169.841508	26.978557

QUESTION 1¶

1.1 Importing the Libraries¶

QUESTION 2¶

2.1 Random RGB and Grayscale Visualization¶

QUESTION 3¶

QUESTION 4¶

QUESTION 5¶

QUESTION 6¶

QUESTION 7¶